refactor: update image generation configuration to remove TPM references and clarify RPM usage in comments

surfsense_backend/app/config/global_llm_config.example.yaml

@@ -208,8 +208,7 @@ global_image_generation_configs:
     model_name: "dall-e-3"
     api_key: "sk-your-openai-api-key-here"
     api_base: ""
-    rpm: 50
-    tpm: 100000
+    rpm: 50  # Requests per minute (image gen is rate-limited by RPM, not tokens)
     litellm_params: {}
 
   # Example: OpenAI GPT Image 1
@@ -221,7 +220,6 @@ global_image_generation_configs:
     api_key: "sk-your-openai-api-key-here"
     api_base: ""
     rpm: 50
-    tpm: 100000
     litellm_params: {}
 
   # Example: Azure OpenAI DALL-E 3
@@ -234,7 +232,6 @@ global_image_generation_configs:
     api_base: "https://your-resource.openai.azure.com"
     api_version: "2024-02-15-preview"
     rpm: 50
-    tpm: 100000
     litellm_params:
       base_model: "dall-e-3"
 
@@ -247,7 +244,6 @@ global_image_generation_configs:
   #   api_key: "your-openrouter-api-key-here"
   #   api_base: ""
   #   rpm: 30
-  #   tpm: 50000
   #   litellm_params: {}
 
 # Notes:
@@ -262,17 +258,11 @@ global_image_generation_configs:
 # - rpm/tpm: Optional rate limits for load balancing (requests/tokens per minute)
 #   These help the router distribute load evenly and avoid rate limit errors
 #
-# AZURE-SPECIFIC NOTES:
-# - Always add 'base_model' in litellm_params for Azure deployments
-# - This fixes "Could not identify azure model 'X'" warnings
-# - base_model should match the underlying OpenAI model (e.g., gpt-4o, gpt-4-turbo, gpt-3.5-turbo)
-# - model_name format: "azure/<your-deployment-name>"
-# - api_version: Use a recent Azure API version (e.g., "2024-02-15-preview")
-# - See: https://docs.litellm.ai/docs/proxy/cost_tracking#spend-tracking-for-azure-openai-models
 #
 # IMAGE GENERATION NOTES:
 # - Image generation configs use the same ID scheme as LLM configs (negative for global)
 # - Supported models: dall-e-2, dall-e-3, gpt-image-1 (OpenAI), azure/* (Azure), 
 #   bedrock/* (AWS), vertex_ai/* (Google), recraft/* (Recraft), openrouter/* (OpenRouter)
 # - The router uses litellm.aimage_generation() for async image generation
-# - api_version is required for Azure image generation deployments
+# - Only RPM (requests per minute) is relevant for image generation rate limiting.
+#   TPM (tokens per minute) does not apply since image APIs are billed/rate-limited per request, not per token.

surfsense_backend/app/services/image_gen_router_service.py

@@ -183,11 +183,11 @@ class ImageGenRouterService:
                 "litellm_params": litellm_params,
             }
 
-            # Add rate limits from config if available
+            # Add RPM rate limit from config if available
+            # Note: TPM (tokens per minute) is not applicable for image generation
+            # since image APIs are rate-limited by requests, not tokens.
             if config.get("rpm"):
                 deployment["rpm"] = config["rpm"]
-            if config.get("tpm"):
-                deployment["tpm"] = config["tpm"]
 
             return deployment
 
@@ -219,10 +219,6 @@ class ImageGenRouterService:
         prompt: str,
         model: str = "auto",
         n: int | None = None,
-        quality: str | None = None,
-        size: str | None = None,
-        style: str | None = None,
-        response_format: str | None = None,
         timeout: int = 600,
         **kwargs,
     ) -> ImageResponse:
@@ -232,16 +228,16 @@ class ImageGenRouterService:
         Uses Router.aimage_generation() which distributes requests
         across configured image generation deployments.
 
+        Parameters like size, quality, style, and response_format are intentionally
+        omitted to keep the interface model-agnostic. Providers use their own
+        sensible defaults. If needed, pass them via **kwargs.
+
         Args:
             prompt: Text description of the desired image(s)
             model: Model alias (default "auto" for router routing)
             n: Number of images to generate
-            quality: Image quality setting
-            size: Image size
-            style: Style parameter
-            response_format: "url" or "b64_json"
             timeout: Request timeout in seconds
-            **kwargs: Additional litellm params
+            **kwargs: Additional provider-specific params (size, quality, etc.)
 
         Returns:
             ImageResponse from litellm
@@ -264,14 +260,6 @@ class ImageGenRouterService:
         }
         if n is not None:
             gen_kwargs["n"] = n
-        if quality is not None:
-            gen_kwargs["quality"] = quality
-        if size is not None:
-            gen_kwargs["size"] = size
-        if style is not None:
-            gen_kwargs["style"] = style
-        if response_format is not None:
-            gen_kwargs["response_format"] = response_format
         gen_kwargs.update(kwargs)
 
         return await instance._router.aimage_generation(**gen_kwargs)